import pandas as pd

import matplotlib.pyplot as plt
from matplotlib import cm
from mpl_toolkits.mplot3d import Axes3D
import seaborn as sns

import ast
import numpy as np

from scipy.stats import chi2_contingency
from scipy.stats import f_oneway, ttest_ind

from sklearn.utils import resample
from sklearn.cluster import KMeans
from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, classification_report, confusion_matrix, roc_curve, auc, roc_auc_score, silhouette_score, silhouette_samples
from sklearn.decomposition import PCA
from sklearn.model_selection import train_test_split

import statsmodels.api as sm
from statsmodels.tools import add_constant

data = pd.read_csv('credit-score-classification-cleaned-dataset/credit_score_cleaned_train.csv')
data.head()

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 100000 entries, 0 to 99999
Data columns (total 28 columns):
 #   Column                    Non-Null Count   Dtype  
---  ------                    --------------   -----  
 0   id                        100000 non-null  object 
 1   customer_id               100000 non-null  object 
 2   month                     100000 non-null  object 
 3   name                      100000 non-null  object 
 4   age                       100000 non-null  int64  
 5   ssn                       100000 non-null  object 
 6   occupation                100000 non-null  object 
 7   annual_income             100000 non-null  float64
 8   monthly_inhand_salary     100000 non-null  float64
 9   total_emi_per_month       100000 non-null  float64
 10  num_bank_accounts         100000 non-null  int64  
 11  num_credit_card           100000 non-null  int64  
 12  interest_rate             100000 non-null  int64  
 13  num_of_loan               100000 non-null  int64  
 14  type_of_loan              100000 non-null  object 
 15  delay_from_due_date       100000 non-null  int64  
 16  num_of_delayed_payment    100000 non-null  int64  
 17  changed_credit_limit      100000 non-null  float64
 18  num_credit_inquiries      100000 non-null  int64  
 19  credit_mix                100000 non-null  object 
 20  outstanding_debt          100000 non-null  float64
 21  credit_utilization_ratio  100000 non-null  float64
 22  credit_history_age        100000 non-null  int64  
 23  payment_of_min_amount     100000 non-null  object 
 24  amount_invested_monthly   100000 non-null  float64
 25  payment_behaviour         100000 non-null  object 
 26  monthly_balance           100000 non-null  float64
 27  credit_score              100000 non-null  int64  
dtypes: float64(8), int64(10), object(10)
memory usage: 21.4+ MB

nullCounts = data.isnull().sum()
print(nullCounts)

id                          0
customer_id                 0
month                       0
name                        0
age                         0
ssn                         0
occupation                  0
annual_income               0
monthly_inhand_salary       0
total_emi_per_month         0
num_bank_accounts           0
num_credit_card             0
interest_rate               0
num_of_loan                 0
type_of_loan                0
delay_from_due_date         0
num_of_delayed_payment      0
changed_credit_limit        0
num_credit_inquiries        0
credit_mix                  0
outstanding_debt            0
credit_utilization_ratio    0
credit_history_age          0
payment_of_min_amount       0
amount_invested_monthly     0
payment_behaviour           0
monthly_balance             0
credit_score                0
dtype: int64

duplicateRows = data[data.duplicated()]
print(f"Number of duplicate rows: {duplicateRows.shape[0]}")

Number of duplicate rows: 0

data.describe().T

data.describe(include="object").T

columnsToDrop = ['id', 'name', 'ssn']
data = data.drop(columns=columnsToDrop)

filteredData = data[data['age'] < 18]
numCustomersToRemove = filteredData['customer_id'].nunique()
print(numCustomersToRemove)

788

customersToRemove = filteredData['customer_id'].unique()
data = data[~data['customer_id'].isin(customersToRemove)]
data['customer_id'].nunique()

11712

filteredData = data[(data['num_bank_accounts'] == 0) & (data['monthly_balance'] > 0)]
numCustomersToRemove = filteredData['customer_id'].nunique()

print(numCustomersToRemove)

563

customersToRemove = filteredData['customer_id'].unique()
data = data[~data['customer_id'].isin(customersToRemove)]
data['customer_id'].nunique()

11149

filteredData = data[(data['credit_history_age']//12) > (data['age']-18)]
print(filteredData[['age', 'credit_history_age']])
numCustomersToRemove = filteredData['customer_id'].nunique()
print(numCustomersToRemove)

       age  credit_history_age
0       23                 265
1       23                 266
2       23                 267
3       23                 268
4       23                 269
...    ...                 ...
99995   25                 378
99996   25                 379
99997   25                 380
99998   25                 381
99999   25                 382

[46537 rows x 2 columns]
5935

data = data.drop('credit_history_age', axis=1)
data.head()

uniqueCreditMix = data['credit_mix'].unique()
print("Unique credit mix in the dataset:", sorted(uniqueCreditMix))

Unique credit mix in the dataset: ['Bad', 'Good', 'Standard']

creditMixMapping = {'Bad': 0, 'Standard': 1, 'Good': 2}
data['creditMixNum'] = data['credit_mix'].map(creditMixMapping)

data[['credit_mix', 'creditMixNum']].head()

uniquePaymentMinAmt = data['payment_of_min_amount'].unique()
print("Unique payment of min amount in the dataset:", sorted(uniquePaymentMinAmt))

Unique payment of min amount in the dataset: ['No', 'Yes']

paymentMinAmtMapping = {'No': 0, 'Yes': 1}
data['paymentMinAmountNum'] = data['payment_of_min_amount'].map(paymentMinAmtMapping)

data[['payment_of_min_amount', 'paymentMinAmountNum']].head()

data['type_of_loan'] = data['type_of_loan'].apply(ast.literal_eval)

def countValidLoans(loanList):
    count = 0
    for loan in loanList:
        if loan != 'No Loan':
            count += 1
    return count

mismatchedRows = data[data['num_of_loan'] != data['type_of_loan'].apply(countValidLoans)]

print(f"Number of rows with mismatched loan counts: {len(mismatchedRows)}")

Number of rows with mismatched loan counts: 0

uniqueLoanTypes = set()

for loans in data['type_of_loan']:
    for loan in loans:  
        uniqueLoanTypes.add(loan) 

loanTypes = list(uniqueLoanTypes)
print(loanTypes)

['No Loan', 'Credit-Builder Loan', 'Personal Loan', 'Payday Loan', 'Mortgage Loan', 'Auto Loan', 'Home Equity Loan', 'Debt Consolidation Loan', 'Student Loan', 'Not Specified']

loanDict = {}

for creditScore in data['credit_score'].unique():
    creditScoreData = data[data['credit_score'] == creditScore]
    loanCounts = {}
    for loans in creditScoreData['type_of_loan']:
        for loan in loans:
            if loan in loanCounts:
                loanCounts[loan] += 1
            else:
                loanCounts[loan] = 1
    loanDict[creditScore] = loanCounts

print(loanDict)

{2: {'Auto Loan': 3566, 'Credit-Builder Loan': 4101, 'Personal Loan': 3659, 'Home Equity Loan': 3802, 'Not Specified': 3993, 'No Loan': 3013, 'Debt Consolidation Loan': 3574, 'Payday Loan': 3897, 'Mortgage Loan': 3845, 'Student Loan': 3888}, 1: {'Auto Loan': 16927, 'Credit-Builder Loan': 18131, 'Personal Loan': 17644, 'Home Equity Loan': 17504, 'Not Specified': 17529, 'No Loan': 6137, 'Student Loan': 16822, 'Debt Consolidation Loan': 17522, 'Payday Loan': 17962, 'Mortgage Loan': 17225}, 0: {'Not Specified': 13374, 'Auto Loan': 13043, 'Student Loan': 13698, 'Personal Loan': 13361, 'Payday Loan': 14101, 'Home Equity Loan': 13286, 'No Loan': 1130, 'Mortgage Loan': 13418, 'Debt Consolidation Loan': 13424, 'Credit-Builder Loan': 13600}}

flattenedData = []

for creditScore, loans in loanDict.items():
    for loan, count in loans.items():
        flattenedData.append({
            'credit_score': creditScore,
            'loan_type': loan,
            'loan_count': count
        })

loanData = pd.DataFrame(flattenedData)
loanData

uniqueScores = data['credit_score'].unique()
print("Unique credit scores in the dataset:", sorted(uniqueScores))

Unique credit scores in the dataset: [0, 1, 2]

scoreCounts = data['credit_score'].value_counts()
scorePercentages = data['credit_score'].value_counts(normalize=True) * 100
scoreSummary = pd.DataFrame({
    'Count': scoreCounts,
    'Percentage (%)': scorePercentages
})

print("Credit score counts and percentages:\n", scoreSummary.sort_index())

Credit score counts and percentages:
    Count  Percentage (%)
0  25361       28.434164
1  48214       54.056418
2  15617       17.509418

labels = ['0 (Poor)', '1 (Standard)', '2 (Good)']
colours = ['#d6e7e4', '#87bbae', '#48756a']
explode = (0, 0.1, 0)
sortedScoreCounts = scoreCounts.sort_index()

plt.figure(figsize=(8, 8))

plt.pie(sortedScoreCounts, 
        autopct=lambda pct: f'{pct:.1f}%', 
        labels=labels,
        startangle=90, 
        colors=colours, 
        textprops={'fontsize': 20},
        wedgeprops={'edgecolor': 'black', 'linewidth': 1.5},
        explode=explode)

plt.title('Distribution of Credit Scores', fontsize=20)
plt.savefig('graphs/credit_score_distribution.png', dpi=300, transparent=True)
plt.show()

uniqueCustomers = data['customer_id'].nunique()
print(f"Number of unique customer IDs: {uniqueCustomers}")

Number of unique customer IDs: 11149

uniqueMonths = data['month'].unique()
print("Unique months in the dataset:", uniqueMonths)

Unique months in the dataset: ['January' 'February' 'March' 'April' 'May' 'June' 'July' 'August']

monthCounts = data['month'].value_counts()

sns.barplot(
    x=monthCounts.values, 
    y=monthCounts.index, 
    color='#498699',
    width=0.5)

plt.xlim(0, 13000)
plt.xlabel("Number of Entries")
plt.ylabel("Month")
plt.title("Number of Entries per Month in the Dataset")
plt.show()

monthOrder = ['August', 'July', 'June', 'May', 'April', 'March', 'February', 'January']
data['orderedMonth'] = pd.Categorical(data['month'], categories=monthOrder)

monthScoreDistribution = data.groupby(['orderedMonth', 'credit_score']).size().unstack()

monthScoreDistribution.plot(kind='barh', stacked=True, color=['#d6e7e4', '#87bbae', '#48756a'])
plt.title("Distribution of Credit Scores per Month")
plt.xlabel("Count")
plt.ylabel("Month")
plt.legend(labels=["0 (Poor)", "1 (Standard)", "2 (Good)"], title="Credit Score", loc="upper right")
plt.show()

creditScoreVariability = data.groupby('customer_id')['credit_score'].nunique()
varyingCreditScores = creditScoreVariability[creditScoreVariability > 1]

print(f"Number of customers with varying credit scores: {len(varyingCreditScores)}")

Number of customers with varying credit scores: 6424

variabilityCounts = creditScoreVariability.value_counts().sort_index()

sns.barplot(
    x=variabilityCounts.values,    
    y=variabilityCounts.index.astype(str),
    color='#498699')

for index, value in enumerate(variabilityCounts.values):
    plt.text(value + 50, index, str(value))
    
plt.title("Credit Score Variability per Customer across months")
plt.ylabel("Number of Unique Credit Scores")
plt.xlabel("Number of Customers")
plt.show()

numericColumns = data.select_dtypes(include='number').columns
numPlots = len(numericColumns)
rows = (numPlots // 2) + 1
cols = 4  
fig, axes = plt.subplots(rows, cols, figsize=(15, rows * 4))
axes = axes.flatten()

for i, col in enumerate(numericColumns):
  
    sns.boxplot(data=data, y=col, ax=axes[2 * i], color="#498699",  width=0.2)
    axes[2 * i].set_title(col)
    axes[2 * i].set_ylabel(col)

    sns.histplot(data=data, x=col, ax=axes[2 * i + 1], color="#98c5e1", kde=True, bins=20)
    mean = data[col].mean()
    axes[2 * i + 1].axvline(mean, color='red', linestyle='--', label=f'Mean: {mean:.2f}')
    axes[2 * i + 1].legend()
    axes[2 * i + 1].set_title(col)
    axes[2 * i + 1].set_xlabel(col)
    axes[2 * i + 1].set_ylabel('Frequency')


for j in range(2 * numPlots, len(axes)):
    fig.delaxes(axes[j])

plt.tight_layout()
plt.show()

columnsToPlot = ['total_emi_per_month', 'amount_invested_monthly']

for column in columnsToPlot:
    Q1 = data[column].quantile(0.25)
    Q3 = data[column].quantile(0.75)
    iqr = Q3 - Q1
    lowerBound = Q1 - 1.5 * iqr
    upperBound = Q3 + 1.5 * iqr
    data = data[(data[column] >= lowerBound) & (data[column] <= upperBound)]

rows = 2
cols = 2  
fig, axes = plt.subplots(rows, cols, figsize=(8, rows * 4))
axes = axes.flatten()


for i, col in enumerate(columnsToPlot):
  
    sns.boxplot(data=data, y=col, ax=axes[2 * i], color="#498699",  width=0.2)
    axes[2 * i].set_title(col)
    axes[2 * i].set_ylabel(col)

    sns.histplot(data=data, x=col, ax=axes[2 * i + 1], color="#98c5e1", kde=True, bins=20)
    mean = data[col].mean()
    axes[2 * i + 1].axvline(mean, color='red', linestyle='--', label=f'Mean: {mean:.2f}')
    axes[2 * i + 1].legend()
    axes[2 * i + 1].set_title(col)
    axes[2 * i + 1].set_xlabel(col)
    axes[2 * i + 1].set_ylabel('Frequency')


for j in range(2 * 2, len(axes)):
    fig.delaxes(axes[j])

plt.tight_layout()
plt.show()

uniqueCustomers = data['customer_id'].nunique()
print(f"Number of unique customer IDs: {uniqueCustomers}")

Number of unique customer IDs: 10720

categoricalColumns = ['month', 'occupation', 'credit_mix', 'payment_behaviour', 'payment_of_min_amount']
dataFiltered = data[categoricalColumns]
numPlots = len(categoricalColumns)
rows = (numPlots // 2) + 1
cols = 2  
fig, axes = plt.subplots(rows, cols, figsize=(12, 10))
axes = axes.flatten()

for i, col in enumerate(categoricalColumns):
    sns.countplot(data=dataFiltered, y=col, ax=axes[i], color='#498699', width=0.5, legend=False, orient='h')
    axes[i].set_title(col)
    axes[i].set_ylabel(col)
    axes[i].set_xlabel('Count')

loanTypeCounts = loanData.groupby('loan_type')['loan_count'].sum().reset_index()
sns.barplot(x='loan_count', y='loan_type', data=loanTypeCounts, ax=axes[numPlots], color='#498699', width=0.5)
axes[numPlots].set_title('type_of_loan')
axes[numPlots].set_xlabel('Count')
axes[numPlots].set_ylabel('type_of_loan')

for j in range(numPlots+1, len(axes)):
    fig.delaxes(axes[j])

plt.tight_layout()
plt.show()

numericColumns = data.select_dtypes(include='number').columns.drop('credit_score')
customPalette = {0: "#d6e7e4", 1: "#87bbae", 2: "#48756a"}

numPlots = len(numericColumns)
nCols = 4
nRows = (numPlots // 2) + 1
fig, axes = plt.subplots(nRows, nCols, figsize=(10, nRows * 4))
axes=axes.flatten()

for i, col in enumerate(numericColumns):
    sns.boxplot(data=data, x='credit_score', y=col, ax=axes[i], hue='credit_score', palette=customPalette, width=0.5, legend=False, flierprops={'markersize': 3})
    axes[i].set_title(col)
    axes[i].set_xlabel('Credit Score')
    axes[i].set_ylabel(col)

for j in range(numPlots, len(axes)):
    axes[j].set_visible(False)

plt.tight_layout()
plt.show()

creditScores = {0: "Credit Score: 0", 1: "Credit Score: 1", 2: "Credit Score: 2"}
customPalette = {0: '#d6e7e4', 1: '#87bbae', 2: '#48756a'}

# Getting the limits for the y-axis for each numeric column by setting it equal to
# the value of the biggest bin in the histogram across all credit scores and 
# dividing it by the sum of all the bins and adding a buffer to leave some space 
# on top

globalXLimits = {}
globalYLimits = {}

for i, col in enumerate(numericColumns):
    xLimit = data[col].max()
    yLimit = 0
    yLimitPercent = 0
    _, bins, _ = plt.hist(data[col], bins=30)
    plt.close()
    for score in creditScores.keys():
        counts = [0] * (len(bins) - 1)
        for i in range(len(bins) - 1):
            creditScoreData = data[data['credit_score'] == score]
            counts[i] = sum((creditScoreData[col] >= bins[i]) & (creditScoreData[col] < bins[i + 1]))
        maxCounts = max(counts)
        if maxCounts > yLimit:
            yLimit = maxCounts
            yLimitPercent = yLimit/sum(counts)

    buffer = yLimitPercent * 0.5
    globalXLimits[col] = (0, xLimit)
    globalYLimits[col] = (0, yLimitPercent + buffer)


nCols = 3 
nRows = len(numericColumns)
fig, axes = plt.subplots(nRows, nCols, figsize=(15, nRows * 4))
axes = axes.reshape(nRows, nCols)

for i, col in enumerate(numericColumns):
    for j, score in enumerate(creditScores.keys()):
        sns.histplot(
            data=data[data['credit_score'] == score],
            x=col,
            kde=True,
            stat="probability",
            color=customPalette[score],
            ax=axes[i, j],
            bins=30
        )
        axes[i, j].set_title(f'{creditScores[score]}: {col}')
        axes[i, j].set_ylabel('Proportion')
        axes[i, j].set_xlabel(col)
        axes[i, j].set_xlim(globalXLimits[col])
        axes[i, j].set_ylim(globalYLimits[col])

plt.tight_layout()
plt.show()

categoricalColumns = ['month', 'occupation', 'credit_mix', 'payment_behaviour', 'payment_of_min_amount']
dataFiltered = data[categoricalColumns]
customPalette = {0: '#d6e7e4', 1: '#87bbae', 2: '#48756a'}

numPlots = len(categoricalColumns)
nCols = 2
nRows = (numPlots + 1) // nCols
fig, axes = plt.subplots(nRows, nCols, figsize=(12, nRows * 6))
axes = axes.flatten()

for i, col in enumerate(categoricalColumns):
    crossTab = pd.crosstab(data[col], data['credit_score'], normalize='index') * 100
    crossTab.plot(kind='barh', color=customPalette, ax=axes[i])  
    axes[i].set_title(f'Distribution of Credit Score by {col}')
    axes[i].set_xlabel('Percentage')
    axes[i].set_ylabel(col)
    axes[i].legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
    axes[i].grid(axis='x', alpha=0.5) 

loanDataPercentage = pd.crosstab(loanData['loan_type'], loanData['credit_score'], loanData['loan_count'], aggfunc='sum', normalize='index') * 100
loanDataPercentage.plot(kind='barh', color=customPalette, ax=axes[numPlots])
axes[numPlots].set_title('Distribution of Credit Score by type_of_loan')
axes[numPlots].set_xlabel('Percentage')
axes[numPlots].set_ylabel('type_of_loan')
axes[numPlots].legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
axes[numPlots].grid(axis='x', alpha=0.5) 


for j in range(numPlots+1, len(axes)):
    axes[j].set_visible(False)

plt.tight_layout()
plt.show()

creditMixDistributionPercent = pd.crosstab(data['credit_score'], data['credit_mix'], normalize='index') * 100
creditMixDistributionPercent = creditMixDistributionPercent[['Bad', 'Standard', 'Good']]   

creditMixDistributionPercent.plot(kind='barh', figsize=(8, 4), color=['#c7dbed', '#81afd3', '#3171a2'])
plt.title('Distribution of Credit Mix in each Credit Score (Percentage)')
plt.ylabel('Credit Score')
plt.xlabel('Percentage')
plt.legend(title='Credit Mix')
plt.xticks(range(0, 101, 5))
plt.grid(axis='x', alpha=0.5) 
plt.tight_layout()
plt.show()

minAmtPaymentDistributionPercent = pd.crosstab(data['credit_score'], data['payment_of_min_amount'], normalize='index') * 100

minAmtPaymentDistributionPercent.plot(kind='barh', figsize=(8, 4), color=['#c7dbed', '#3171a2'])
plt.title('Distribution of Payment of Just Minimum Amount in each Credit Score (Percentage)')
plt.ylabel('Credit Score')
plt.xlabel('Percentage')
plt.legend(title='Payment of Just Minimum Amount')
plt.xticks(range(0, 101, 5))
plt.grid(axis='x', alpha=0.5) 
plt.tight_layout()
plt.show()

correlationMatrix = data.corr(numeric_only=True)

plt.figure(figsize=(12, 8))
sns.heatmap(correlationMatrix, annot=True, cmap=sns.diverging_palette(220, 20, as_cmap=True), fmt='.2f', cbar=True, 
            xticklabels=correlationMatrix.columns, yticklabels=correlationMatrix.columns)

plt.title('Correlation Matrix Heatmap')
plt.show()

correlationWithCreditScore = correlationMatrix['credit_score'].sort_values(key=lambda x: x.abs(), ascending=False)

correlationTable = correlationWithCreditScore.reset_index()
correlationTable.columns = ['Column', 'Correlation with Credit Score']
correlationTableDF = pd.DataFrame(correlationTable)
correlationTableDF

def correlation_ratio(categories, values):     # (Kiryl, 2021)                       
    categories = np.array(categories)                                
    values = np.array(values)                                        
    
    ssw = 0
    ssb = 0
    for category in set(categories):
        subgroup = values[np.where(categories == category)[0]]
        ssw += sum((subgroup-np.mean(subgroup))**2)
        ssb += len(subgroup)*(np.mean(subgroup)-np.mean(values))**2

    return (ssb / (ssb + ssw))

results = {}
scores = [0, 1, 2]
for column in numericColumns:
    if column != 'credit_score':
        groupData = []
        for score in scores:
            columnData = data[data['credit_score'] == score][column].tolist()
            groupData.append(columnData)
            
        f_stat, p_value = f_oneway(groupData[0], groupData[1], groupData[2])
        results[column] = {'F-statistic': f_stat, 'p-value': p_value}

        if (p_value <= 0.05):
            etaSquared = correlation_ratio(data['credit_score'], data[column])
            results[column]['etaSquared'] = etaSquared


resultsDF = pd.DataFrame(results).T
resultsDF.sort_values('etaSquared', ascending=False, inplace=True)

resultsDF

def cramersCorrectedStat(confusionMatrix):      # (Eunicien & Lee, 2021)                                     
    """ calculate Cramers V statistic for categorial-categorial association.                
        uses correction from Bergsma and Wicher, 
        Journal of the Korean Statistical Society 42 (2013): 323-328
    """
    chi2 = chi2_contingency(confusionMatrix)[0]
    n = confusionMatrix.sum().sum()
    phi2 = chi2/n
    r,k = confusionMatrix.shape
    phi2corr = max(0, phi2 - ((k-1)*(r-1))/(n-1))    
    rcorr = r - ((r-1)**2)/(n-1)
    kcorr = k - ((k-1)**2)/(n-1)
    return np.sqrt(phi2corr / min( (kcorr-1), (rcorr-1)))

results = {}

for col in categoricalColumns:
    contingencyTable = pd.crosstab(data['credit_score'], data[col])
    chi2, p, _, _ = chi2_contingency(contingencyTable)
    results[col] = {'chi2': chi2, 'p-value': p}

    if (p <= 0.05):
        results[col]["cramer's v"] = cramersCorrectedStat(contingencyTable)


contingencyTable = pd.crosstab(loanData['credit_score'], loanData['loan_type'], loanData['loan_count'], aggfunc='sum')
chi2, p, _, _ = chi2_contingency(contingencyTable)
results['type_of_loan'] = {'chi2': chi2, 'p-value': p}

if (p <= 0.05):
    results['type_of_loan']["cramer's v"] = cramersCorrectedStat(contingencyTable)    

resultsDF = pd.DataFrame(results).T
resultsDF.sort_values("cramer's v", ascending=False, inplace=True)
resultsDF

customerData = data.groupby('customer_id').agg({
    'outstanding_debt': 'mean',
    'num_of_loan': 'mean',
    'num_credit_card': 'mean',
    'delay_from_due_date': 'mean',
    'num_of_delayed_payment': 'mean',
    'paymentMinAmountNum': 'mean',
    'age': 'mean',
    'annual_income': 'mean',
    'credit_score': 'mean'
}).reset_index()

customerData['num_of_loan'] = customerData['num_of_loan'].round().astype(int)
customerData['num_credit_card'] = customerData['num_credit_card'].round().astype(int)
customerData['delay_from_due_date'] = customerData['delay_from_due_date'].round().astype(int)
customerData['num_of_delayed_payment'] = customerData['num_of_delayed_payment'].round().astype(int)
customerData['paymentMinAmountNum'] = customerData['paymentMinAmountNum'].round().astype(int)
customerData['age'] = customerData['age'].round().astype(int)
customerData['annual_income'] = customerData['annual_income'].round().astype(int)

customerData['credit_score'] = customerData['credit_score'].round().astype(int)

customerData.head()

scoreCounts = customerData['credit_score'].value_counts()
scorePercentages = customerData['credit_score'].value_counts(normalize=True) * 100
scoreSummary = pd.DataFrame({
    'Count': scoreCounts,
    'Percentage (%)': scorePercentages
})


print("Credit score counts and percentages:\n", scoreSummary.sort_index())

Credit score counts and percentages:
    Count  Percentage (%)
0   3512       32.761194
1   5231       48.796642
2   1977       18.442164

plt.pie(scoreCounts.sort_index(), labels=['0 (Poor)', '1 (Standard)', '(2) Good'], autopct='%1.1f%%', colors=['#d6e7e4', '#87bbae', '#48756a'])
plt.title("Distribution of Credit Scores")
plt.axis('equal')
plt.show()

correlationMatrix = customerData.corr(numeric_only=True)

plt.figure(figsize=(12, 8))
sns.heatmap(correlationMatrix, annot=True, cmap=sns.diverging_palette(220, 20, as_cmap=True), fmt='.2f', cbar=True, 
            xticklabels=correlationMatrix.columns, yticklabels=correlationMatrix.columns)

plt.title('Correlation Matrix Heatmap')
plt.show()

debtManagementColumns = ['outstanding_debt', 'num_of_loan', 'num_credit_card']

numPlots = len(debtManagementColumns)
rows = (numPlots // 2) + 1
cols = 4  

fig, axes = plt.subplots(rows, cols, figsize=(15, rows * 4))
axes = axes.flatten()

for i, col in enumerate(debtManagementColumns):
    sns.boxplot(data=customerData, y=col, ax=axes[2 * i], color="#498699",  width=0.2)
    axes[2 * i].set_title(col)
    axes[2 * i].set_ylabel(col)

    sns.histplot(data=data, x=col, ax=axes[2 * i + 1], color="#98c5e1", kde=True, bins=20)
    mean_value = data[col].mean()
    axes[2 * i + 1].axvline(mean_value, color='red', linestyle='--', label=f'Mean: {mean_value:.2f}')
    axes[2 * i + 1].legend()

    axes[2 * i + 1].set_title(col)
    axes[2 * i + 1].set_xlabel(col)
    axes[2 * i + 1].set_ylabel('Frequency')

for j in range(2 * numPlots, len(axes)):
    fig.delaxes(axes[j])

plt.tight_layout()
plt.show()

customPalette = {0: "#d6e7e4", 1: "#87bbae", 2: "#48756a"}

fig, axes = plt.subplots(rows, cols, figsize=(8, rows * 4))
axes = axes.flatten()

for i, col in enumerate(debtManagementColumns):
    sns.boxplot(data=customerData, x='credit_score', y=col, ax=axes[i], hue='credit_score', palette=customPalette, width=0.5, legend=False, flierprops={'markersize': 3})
    axes[i].set_title(col)
    axes[i].set_xlabel('Credit Score')
    axes[i].set_ylabel(col)

for j in range(numPlots, len(axes)):
    axes[j].set_visible(False)

plt.tight_layout()
plt.savefig('graphs/debtManagement(EDA).png', dpi=300, transparent=True)
plt.show()

def correlation_ratio(categories, values):   # (Kiryl, 2021)
    categories = np.array(categories)
    values = np.array(values)
    
    ssw = 0
    ssb = 0
    for category in set(categories):
        subgroup = values[np.where(categories == category)[0]]
        ssw += sum((subgroup-np.mean(subgroup))**2)
        ssb += len(subgroup)*(np.mean(subgroup)-np.mean(values))**2

    return (ssb / (ssb + ssw))

results = {}
scores = [0, 1, 2]
for column in debtManagementColumns:
    groupData = []
    for score in scores:
        columnData = customerData[customerData['credit_score'] == score][column].tolist()
        groupData.append(columnData)
        
    f_stat, p_value = f_oneway(groupData[0], groupData[1], groupData[2])
    results[column] = {'F-statistic': f_stat, 'p-value': p_value}

    if (p_value <= 0.05):
        etaSquared = correlation_ratio(customerData['credit_score'], customerData[column])
        results[column]['etaSquared'] = etaSquared


resultsDF = pd.DataFrame(results).T
resultsDF.sort_values('etaSquared', ascending=False, inplace=True)
resultsDF

features = ['outstanding_debt', 'num_of_loan']
scaler = StandardScaler()
debtManagementScaled = scaler.fit_transform(customerData[features])

clusterRange = [2,3,4,5,6] 
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):                                            
    ax = axes[i]    
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(debtManagementScaled)
    silhouetteAvg = silhouette_score(debtManagementScaled, clusterLabels)
    silhouetteValues = silhouette_samples(debtManagementScaled, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.5588
Number of Clusters: 3, Silhouette Score: 0.4256
Number of Clusters: 4, Silhouette Score: 0.3821
Number of Clusters: 5, Silhouette Score: 0.4039
Number of Clusters: 6, Silhouette Score: 0.3934

dataClusters = customerData.copy()
kmeans = KMeans(n_clusters=5, random_state=42, n_init='auto')
dataClusters['debt_management_cluster'] = kmeans.fit_predict(debtManagementScaled)
customLabels = {
    0: "Moderate Debt and Moderate Loans",
    1: "Moderate Debt and Many Loans",
    2: "High Debt and Many Loans",
    3: "Low Debt and Few Loans",
    4: "Low Debt and Moderate Loans",
}
dataClusters['debt_management_cluster'] = dataClusters['debt_management_cluster'].map(customLabels)
plt.figure(figsize=(8, 6))
sns.scatterplot(x=dataClusters['outstanding_debt'], y=dataClusters['num_of_loan'], hue=dataClusters['debt_management_cluster'], palette='deep')
plt.title('Debt Management Clusters')
plt.xlabel('Outstanding Debt')
plt.ylabel('Number of Loans')
plt.grid(alpha=0.5) 
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['debt_management_cluster'], dataClusters['credit_score'], normalize='index') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Debt Management Clusters')
plt.xlabel('Percentage')
plt.ylabel('Debt Management Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(alpha=0.5) 
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(10, 3))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5)

plt.title('Credit Score Distributions In Debt Management Clusters')
plt.ylabel('Debt Management Cluster')
plt.xlabel('Credit Score')
plt.show()

features = ['outstanding_debt', 'num_credit_card']
scaler = StandardScaler()
debtManagementScaled = scaler.fit_transform(customerData[features])

clusterRange = [2,3,4,5,6]
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):
    ax = axes[i]
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(debtManagementScaled)
    silhouetteAvg = silhouette_score(debtManagementScaled, clusterLabels)
    silhouetteValues = silhouette_samples(debtManagementScaled, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.5025
Number of Clusters: 3, Silhouette Score: 0.3993
Number of Clusters: 4, Silhouette Score: 0.3965
Number of Clusters: 5, Silhouette Score: 0.4313
Number of Clusters: 6, Silhouette Score: 0.3682

kmeans = KMeans(n_clusters=5, random_state=42, n_init='auto')
dataClusters['debt_management_cluster'] = kmeans.fit_predict(debtManagementScaled)
customLabels = {
    0: "Low Debt and Moderate Credit Cards",
    1: "High Debt and Many Credit Cards",
    2: "Moderate Debt and Moderate Credit Cards",
    3: "Low Debt and Few Credit Cards",
    4: "Moderate Debt and Many Credit Cards",
}
dataClusters['debt_management_cluster'] = dataClusters['debt_management_cluster'].map(customLabels)

plt.figure(figsize=(8, 6))
sns.scatterplot(x=dataClusters['outstanding_debt'], y=dataClusters['num_credit_card'], hue=dataClusters['debt_management_cluster'], palette='deep')
plt.title('Debt Management Clusters')
plt.xlabel('Outstanding Debt')
plt.ylabel('Number of Credit Cards')
plt.grid(alpha=0.5) 
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['debt_management_cluster'], dataClusters['credit_score'], normalize='index') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Debt Management Clusters')
plt.xlabel('Percentage')
plt.ylabel('Debt Management Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(alpha=0.5) 
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(10, 3))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5)

plt.title('Credit Score Distributions In Debt Management Clusters')
plt.xlabel('Debt Management Cluster')
plt.ylabel('Credit Score')
plt.show()

features = ['outstanding_debt', 'num_of_loan', 'num_credit_card']
scaler = StandardScaler()
debtManagementScaled = scaler.fit_transform(customerData[features])

clusterRange = [2,3,4,5,6] 
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):
    ax = axes[i]
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(debtManagementScaled)
    silhouetteAvg = silhouette_score(debtManagementScaled, clusterLabels)
    silhouetteValues = silhouette_samples(debtManagementScaled, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.4691
Number of Clusters: 3, Silhouette Score: 0.3028
Number of Clusters: 4, Silhouette Score: 0.2907
Number of Clusters: 5, Silhouette Score: 0.2752
Number of Clusters: 6, Silhouette Score: 0.2849

kmeans = KMeans(n_clusters=5, random_state=42, n_init='auto')                                     
dataClusters = customerData.copy()                                                                
dataClusters['debt_management_cluster'] = kmeans.fit_predict(debtManagementScaled)
customLabels = {
    0: "Low Debt, Few Loans, Moderate Credit Cards",
    1: "High Debt, Many Loans, Many Credit Cards",
    2: "Moderate Debt, Many Loans, Many Credit Cards",
    3: "Low Debt, Few Loans, Few Credit Cards",
    4: "Low Debt, Many Loans, Few Credit Cards",
}
dataClusters['debt_management_cluster'] = dataClusters['debt_management_cluster'].map(customLabels)

colours = cm.viridis(np.linspace(0, 1, 5))
fig = plt.figure(figsize=(10, 8))
ax = fig.add_subplot(projection='3d')

for cluster in range(5):
    clusterData = dataClusters[dataClusters['debt_management_cluster'] == customLabels[cluster]]
    ax.scatter(
        clusterData['outstanding_debt'], 
        clusterData['num_of_loan'], 
        clusterData['num_credit_card'], 
        color=colours[cluster], 
        label=f'{customLabels[cluster]}', 
        s=20
    )

ax.set_title('Debt Management Clusters', fontsize=14)
ax.set_xlabel('Outstanding Debt', fontsize=12)
ax.set_ylabel('Number of Loans', fontsize=12)
ax.set_zlabel('Number of Credit Cards', fontsize=12, labelpad=-26)

ax.legend(title='Clusters', loc='upper left', fontsize=10)
plt.savefig('graphs/debtManagementClusters.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['debt_management_cluster'], dataClusters['credit_score'], normalize='index') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Debt Management Clusters')
plt.xlabel('Percentage')
plt.ylabel('Debt Management Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(alpha=0.5)
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(8, 4))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5, annot_kws={"size": 14})

plt.title('Credit Score Distributions In Debt Management Clusters', fontsize=16)
plt.ylabel('Debt Management Cluster')
plt.xlabel('Credit Score')
plt.yticks(fontsize=14)  
plt.xticks(fontsize=14)
plt.savefig('graphs/debtManagementResults.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

x = customerData[['outstanding_debt', 'num_credit_card', 'num_of_loan']]
y = customerData['credit_score']

xTrain, xTest, yTrain, yTest = train_test_split(x, y, test_size=0.2, random_state=42)

scaler = StandardScaler()
xTrain = scaler.fit_transform(xTrain)
xTest = scaler.transform(xTest)

model = LogisticRegression(multi_class='multinomial')
result = model.fit(xTrain, yTrain)

yPred = model.predict(xTest)
accuracy = accuracy_score(yTest, yPred)
print("Accuracy: {:.2f}%".format(accuracy * 100))

Accuracy: 62.03%

classNames = [f"Credit Score {className}" for className in model.classes_]

coefficients_df = pd.DataFrame(
    model.coef_.T, 
    index=x.columns, 
    columns= classNames)

print(coefficients_df)

                  Credit Score 0  Credit Score 1  Credit Score 2
outstanding_debt        0.427311        0.005169       -0.432480
num_credit_card         0.529013        0.079216       -0.608229
num_of_loan             0.293723        0.036926       -0.330649

confusionMatrix = confusion_matrix(yTest, yPred)

plt.figure(figsize=(6, 4))
sns.heatmap(confusionMatrix, annot=True, fmt='d', cmap='Blues', xticklabels=model.classes_, yticklabels=model.classes_)
plt.title("Confusion Matrix")
plt.xlabel('Predicted')
plt.ylabel('True')
plt.show()

paymentBehaviourColumns = ['delay_from_due_date', 'num_of_delayed_payment']

numPlots = len(paymentBehaviourColumns)
rows = 2
cols = 2

fig, axes = plt.subplots(rows, cols, figsize=(10, rows * 4))
axes = axes.flatten()

for i, col in enumerate(paymentBehaviourColumns):
    sns.boxplot(data=customerData, y=col, ax=axes[2 * i], color="#498699",  width=0.2)
    axes[2 * i].set_title(col)
    axes[2 * i].set_ylabel(col)

    sns.histplot(data=data, x=col, ax=axes[2 * i + 1], color="#98c5e1", kde=True, bins=30)
    mean_value = data[col].mean()
    axes[2 * i + 1].axvline(mean_value, color='red', linestyle='--', label=f'Mean: {mean_value:.2f}')
    axes[2 * i + 1].legend()

    axes[2 * i + 1].set_title(col)
    axes[2 * i + 1].set_xlabel(col)
    axes[2 * i + 1].set_ylabel('Frequency')

for j in range(2 * numPlots, len(axes)):
    fig.delaxes(axes[j])

plt.tight_layout()
plt.show()

plt.figure(figsize=(8, 3))
sns.countplot(data=customerData, y='paymentMinAmountNum', color='#498699', width=0.2, legend=False, orient='h')
plt.title('Bar Plot of paymentMinAmountNum')
plt.ylabel('paymentMinAmountNum')
plt.xlabel('Count')

Text(0.5, 0, 'Count')

customPalette = {0: "#d6e7e4", 1: "#87bbae", 2: "#48756a"}

fig, axes = plt.subplots(1, 2, figsize=(5, 5))
axes = axes.flatten()

for i, col in enumerate(paymentBehaviourColumns):
    sns.boxplot(data=customerData, x='credit_score', y=col, ax=axes[i], hue='credit_score', palette=customPalette, width=0.5, legend=False, flierprops={'markersize': 3})
    axes[i].set_title(col)
    axes[i].set_xlabel('Credit Score')
    axes[i].set_ylabel(col)

for j in range(numPlots, len(axes)):
    axes[j].set_visible(False)

plt.tight_layout()
plt.savefig('graphs/paymentBehaviour(EDA).png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

paymentMinAmountDistribution = pd.crosstab(customerData['paymentMinAmountNum'], customerData['credit_score'], normalize='index') * 100

paymentMinAmountDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distribution Across PaymentMinAmountNum')
plt.xlabel('Percentage')
plt.ylabel('PaymentMinAmountNum')
plt.grid(axis='x', alpha=0.5) 
plt.savefig('graphs/paymentBehaviour(EDA)v2.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

results = {}
scores = [0, 1, 2]
for column in paymentBehaviourColumns:
    groupData = []
    for score in scores:
        columnData = customerData[customerData['credit_score'] == score][column].tolist()
        groupData.append(columnData)
        
    f_stat, p_value = f_oneway(groupData[0], groupData[1], groupData[2])
    results[column] = {'F-statistic': f_stat, 'p-value': p_value}

    if (p_value <= 0.05):
        etaSquared = correlation_ratio(customerData['credit_score'], customerData[column])
        results[column]['etaSquared'] = etaSquared


resultsDF = pd.DataFrame(results).T
resultsDF.sort_values('etaSquared', ascending=False, inplace=True)
resultsDF

results = {}
contingencyTable = pd.crosstab(customerData['credit_score'], customerData['paymentMinAmountNum'])
chi2, p, _, _ = chi2_contingency(contingencyTable)
results['paymentMinAmountNum'] = {'chi2': chi2, 'p-value': p}

if (p <= 0.05):
    results['paymentMinAmountNum']["cramer's v"] = cramersCorrectedStat(contingencyTable)

print(results)

{'paymentMinAmountNum': {'chi2': 2879.7318747499944, 'p-value': 0.0, "cramer's v": 0.5181410680112958}}

features = ['delay_from_due_date', 'num_of_delayed_payment']
scaler = StandardScaler()
paymentBehaviourScaled = scaler.fit_transform(customerData[features])

clusterRange = [2,3,4,5,6] 
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):
    ax = axes[i]
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(paymentBehaviourScaled)
    silhouetteAvg = silhouette_score(paymentBehaviourScaled, clusterLabels)
    silhouetteValues = silhouette_samples(paymentBehaviourScaled, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.4218
Number of Clusters: 3, Silhouette Score: 0.4275
Number of Clusters: 4, Silhouette Score: 0.3754
Number of Clusters: 5, Silhouette Score: 0.3993
Number of Clusters: 6, Silhouette Score: 0.3959

kmeans = KMeans(n_clusters=3, random_state=42, n_init='auto')
dataClusters['payment_behaviour_cluster'] = kmeans.fit_predict(paymentBehaviourScaled)
customLabels = {
    0: "Short Delay and Few Delayed Payments",
    1: "Long Delay and Many Delayed Payemnts",
    2: "Short Delay and Many Delayed Payments"
}
dataClusters['payment_behaviour_cluster'] = dataClusters['payment_behaviour_cluster'].map(customLabels)

plt.figure(figsize=(8, 6))
sns.scatterplot(x=dataClusters['delay_from_due_date'], y=dataClusters['num_of_delayed_payment'], hue=dataClusters['payment_behaviour_cluster'], palette='deep')
plt.title('Payment Behaviour Clusters')
plt.xlabel('Delay from due date')
plt.ylabel('Number of delayed payments')
plt.grid(alpha=0.5) 
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['payment_behaviour_cluster'], dataClusters['credit_score'], normalize='index') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Payment Behaviour Clusters')
plt.xlabel('Percentage')
plt.ylabel('Payment Behaviour Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(axis='x', alpha=0.5) 
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(10, 3))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5)

plt.title('Credit Score Distributions In Payment Behaviour Clusters')
plt.xlabel('Payment Behaviour Cluster')
plt.ylabel('Credit Score')
plt.show()

features = ['delay_from_due_date', 'num_of_delayed_payment', 'paymentMinAmountNum']
scaler = StandardScaler()
paymentBehaviourScaled = scaler.fit_transform(customerData[features])

clusterRange = [2,3,4,5,6]
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):
    ax = axes[i]
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(paymentBehaviourScaled)
    silhouetteAvg = silhouette_score(paymentBehaviourScaled, clusterLabels)
    silhouetteValues = silhouette_samples(paymentBehaviourScaled, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.4929
Number of Clusters: 3, Silhouette Score: 0.4794
Number of Clusters: 4, Silhouette Score: 0.4212
Number of Clusters: 5, Silhouette Score: 0.4136
Number of Clusters: 6, Silhouette Score: 0.3998

kmeans = KMeans(n_clusters=5, random_state=42, n_init='auto')
dataClusters['payment_behaviour_cluster'] = kmeans.fit_predict(paymentBehaviourScaled)
customLabels = {
    0: "Short Delay, Many Delayed Payments, Paid more than Min Amount",
    1: "Short Delay, Many Delayed Payments, Paid Just Min Amount",
    2: "Short Delay, Few Delayed Payments, Paid Just Min Amount",
    3: "Long Delay, Many Delayed Payments, Paid Just Min Amount",
    4: "Short Delay, Few Delayed Payments, Paid more than Min Amount"
}
dataClusters['payment_behaviour_cluster'] = dataClusters['payment_behaviour_cluster'].map(customLabels)

colours = cm.viridis(np.linspace(0, 1, 5))
fig = plt.figure(figsize=(10, 8))
ax = fig.add_subplot(projection='3d')

for cluster in range(5):
    clusterData = dataClusters[dataClusters['payment_behaviour_cluster'] == customLabels[cluster]]
    ax.scatter(
        clusterData['delay_from_due_date'], 
        clusterData['num_of_delayed_payment'], 
        clusterData['paymentMinAmountNum'], 
        color=colours[cluster], 
        label=customLabels[cluster], 
        s=20
    )

ax.set_title('Payment Behaviour Clusters', fontsize=16)
ax.set_xlabel('Delay from Due Date', fontsize=14)
ax.set_ylabel('Number of Delayed Payments', fontsize=14)
ax.set_zlabel('Payment of Just Minimum Amount', fontsize=14, labelpad=-30)
ax.tick_params(axis='x', labelsize=12)
ax.tick_params(axis='y', labelsize=12)
ax.tick_params(axis='z', labelsize=12)
ax.legend(title='Clusters', loc='upper left', title_fontsize=12, fontsize=10)
plt.tight_layout()
plt.savefig('graphs/paymentBehaviourClusters.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['payment_behaviour_cluster'], dataClusters['credit_score'], normalize='index') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Payment Behaviour Clusters')
plt.xlabel('Percentage')
plt.ylabel('Payment Behaviour Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(axis='x', alpha=0.5) 
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(8, 3))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5, annot_kws={"size": 16})

plt.title('Credit Score Distributions In Payment Behaviour Clusters', fontsize=16)
plt.ylabel('Payment Behaviour Cluster', fontsize=14)
plt.xlabel('Credit Score', fontsize=14)
plt.xticks(fontsize=14)  
plt.yticks(fontsize=14)
plt.savefig('graphs/paymentBehaviourResults.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

x = customerData[['delay_from_due_date', 'num_of_delayed_payment', 'paymentMinAmountNum']]
y = customerData['credit_score']

xTrain, xTest, yTrain, yTest = train_test_split(x, y, test_size=0.2, random_state=42)

scaler = StandardScaler()
xTrain = scaler.fit_transform(xTrain)
xTest = scaler.transform(xTest)

model = LogisticRegression(multi_class='multinomial')
result = model.fit(xTrain, yTrain)

yPred = model.predict(xTest)
accuracy = accuracy_score(yTest, yPred)
print("Accuracy: {:.2f}%".format(accuracy * 100))

Accuracy: 63.15%

classNames = [f"Credit Score {className}" for className in model.classes_]

coefficients_df = pd.DataFrame(
    model.coef_.T, 
    index=x.columns, 
    columns= classNames)

print(coefficients_df)

                        Credit Score 0  Credit Score 1  Credit Score 2
delay_from_due_date           0.688864       -0.018321       -0.670543
num_of_delayed_payment        0.119488        0.079481       -0.198969
paymentMinAmountNum           0.448782        0.345197       -0.793979

demographicFactorsColumns = ['age', 'annual_income']

numPlots = len(demographicFactorsColumns)
rows = 2
cols = 2

fig, axes = plt.subplots(rows, cols, figsize=(10, rows * 4))
axes = axes.flatten()

for i, col in enumerate(demographicFactorsColumns):
    sns.boxplot(data=customerData, y=col, ax=axes[2 * i], color="#498699",  width=0.2)
    axes[2 * i].set_title(col)
    axes[2 * i].set_ylabel(col)

    sns.histplot(data=data, x=col, ax=axes[2 * i + 1], color="#98c5e1", kde=True, bins=30)
    mean_value = data[col].mean()
    axes[2 * i + 1].axvline(mean_value, color='red', linestyle='--', label=f'Mean: {mean_value:.2f}')
    axes[2 * i + 1].legend()

    axes[2 * i + 1].set_title(col)
    axes[2 * i + 1].set_xlabel(col)
    axes[2 * i + 1].set_ylabel('Frequency')

for j in range(2 * numPlots, len(axes)):
    fig.delaxes(axes[j])

plt.tight_layout()
plt.show()

customPalette = {0: "#d6e7e4", 1: "#87bbae", 2: "#48756a"}
flierprops = dict(markersize=3)

fig, axes = plt.subplots(1, 2, figsize=(5, 5))
axes = axes.flatten()

for i, col in enumerate(demographicFactorsColumns):
    sns.boxplot(data=customerData, x='credit_score', y=col, ax=axes[i], hue='credit_score', palette=customPalette, width=0.5, legend=False, flierprops=flierprops)
    axes[i].set_title(col)
    axes[i].set_xlabel('Credit Score')
    axes[i].set_ylabel(col)

for j in range(numPlots, len(axes)):
    axes[j].set_visible(False)

plt.tight_layout()
plt.savefig('graphs/demographicFactors(EDA).png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

results = {}
scores =[0, 1, 2]

for column in demographicFactorsColumns:
    groupData = []
    for score in scores:
        columnData = customerData[customerData['credit_score'] == score][column].tolist()
        groupData.append(columnData)
        
    f_stat, p_value = f_oneway(groupData[0], groupData[1], groupData[2])
    results[column] = {'F-statistic': f_stat, 'p-value': p_value}

    if (p_value <= 0.05):
        etaSquared = correlation_ratio(customerData['credit_score'], customerData[column])
        results[column]['etaSquared'] = etaSquared


resultsDF = pd.DataFrame(results).T
resultsDF.sort_values('etaSquared', ascending=False, inplace=True)

resultsDF

features = ['age', 'annual_income']
scaler = StandardScaler()
demographicFactorsScaled = scaler.fit_transform(customerData[features])

clusterRange = [2,3,4,5,6]
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):
    ax = axes[i]
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(demographicFactorsScaled)
    silhouetteAvg = silhouette_score(demographicFactorsScaled, clusterLabels)
    silhouetteValues = silhouette_samples(demographicFactorsScaled, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.3629
Number of Clusters: 3, Silhouette Score: 0.4190
Number of Clusters: 4, Silhouette Score: 0.3617
Number of Clusters: 5, Silhouette Score: 0.3673
Number of Clusters: 6, Silhouette Score: 0.3768

kmeans = KMeans(n_clusters=3, random_state=42, n_init='auto')
dataClusters['demographic_factors_cluster'] = kmeans.fit_predict(demographicFactorsScaled)
customLabels = {
    0: "Old, Low Annual Income",
    1: "Young, Low Annual Income",
    2: "High Annual Income"
}
dataClusters['demographic_factors_cluster'] = dataClusters['demographic_factors_cluster'].map(customLabels)

plt.figure(figsize=(8, 6))
sns.scatterplot(x=dataClusters['age'], y=dataClusters['annual_income'], hue=dataClusters['demographic_factors_cluster'], palette='deep')
plt.title('Demographic Factors Clusters', fontsize = 14)
plt.xlabel('Age', fontsize=14)
plt.ylabel('Annual Income', fontsize=14)
plt.xticks(fontsize = 12)
plt.xticks(fontsize = 12)
plt.grid(True, linestyle='--', alpha=0.5) 
plt.savefig('graphs/demographicFactorsClusters.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['demographic_factors_cluster'], dataClusters['credit_score'], normalize='index') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Demographic Factors Clusters')
plt.xlabel('Percentage')
plt.ylabel('Demographic Factors Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(axis='x', alpha=0.5) 
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(10, 3))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5, annot_kws={"size": 16})

plt.title('Credit Score Distributions In Demographic Factors Clusters', fontsize=16)
plt.ylabel('Demographic Factors Cluster', fontsize=14)
plt.xlabel('Credit Score', fontsize=14)
plt.xticks(fontsize=14)  
plt.yticks(fontsize=14)
plt.savefig('graphs/demographicFactorsResults.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

x = customerData[['age', 'annual_income']]
y = customerData['credit_score']

xTrain, xTest, yTrain, yTest = train_test_split(x, y, test_size=0.2, random_state=42)

scaler = StandardScaler()
xTrain = scaler.fit_transform(xTrain)
xTest = scaler.transform(xTest)

model = LogisticRegression(multi_class='multinomial')
result = model.fit(xTrain, yTrain)

yPred = model.predict(xTest)
accuracy = accuracy_score(yTest, yPred)
print("Accuracy: {:.2f}%".format(accuracy * 100))

Accuracy: 49.35%

classNames = [f"Credit Score {className}" for className in model.classes_]

coefficients_df = pd.DataFrame(
    model.coef_.T, 
    index=x.columns, 
    columns= classNames)

print(coefficients_df)

               Credit Score 0  Credit Score 1  Credit Score 2
age                 -0.185097        0.002644        0.182454
annual_income       -0.284361        0.028903        0.255458

scaler = StandardScaler()
debtManagementScaled = scaler.fit_transform(customerData[['outstanding_debt', 'num_of_loan', 'num_credit_card']])

pca = PCA(n_components=1)
debtManagementComponent = pca.fit_transform(debtManagementScaled)
customerData['debtManagementComponent'] = debtManagementComponent
print(f"Explained Variance Ratio: {pca.explained_variance_ratio_}")
print(f"PCA Loadings: {pca.components_}")
customerData.head()

Explained Variance Ratio: [0.67607054]
PCA Loadings: [[0.6130833  0.59114397 0.524097  ]]

scaler = StandardScaler()
paymentBehaviourScaled = scaler.fit_transform(customerData[['delay_from_due_date', 'num_of_delayed_payment', 'paymentMinAmountNum']])

pca = PCA(n_components=1)
paymentBehaviourComponent = pca.fit_transform(paymentBehaviourScaled)
customerData['paymentBehaviourComponent'] = paymentBehaviourComponent
print(f"Explained Variance Ratio: {pca.explained_variance_ratio_}")
print(f"PCA Loadings: {pca.components_}")
customerData.head()

Explained Variance Ratio: [0.69227479]
PCA Loadings: [[0.56370799 0.59153227 0.57647452]]

scaler = StandardScaler()
demographicFactorsScaled = scaler.fit_transform(customerData[['age', 'annual_income']])

pca = PCA(n_components=1)
demographicFactorsComponent = pca.fit_transform(demographicFactorsScaled)
customerData['demographicFactorsComponent'] = demographicFactorsComponent
print(f"Explained Variance Ratio: {pca.explained_variance_ratio_}")
print(f"PCA Loadings: {pca.components_}")
customerData.head()

Explained Variance Ratio: [0.53652232]
PCA Loadings: [[0.70710678 0.70710678]]

customerData['demographicFactorsComponent'] = (
    0.5 * demographicFactorsScaled[:, 0] + 
    0.5 * demographicFactorsScaled[:, 1]
)

customerData.head()

features = customerData[['debtManagementComponent', 'paymentBehaviourComponent', 'demographicFactorsComponent']]

clusterRange = [2,3,4,5,6]
fig, axes = plt.subplots(1, 5, figsize=(25, 5))
fig.suptitle("Silhouette Analysis for K-Means Clustering")

for i, nClusters in enumerate(clusterRange):
    ax = axes[i]
    kmeans = KMeans(n_clusters=nClusters, random_state=42, n_init='auto')
    clusterLabels = kmeans.fit_predict(features)
    silhouetteAvg = silhouette_score(features, clusterLabels)
    silhouetteValues = silhouette_samples(features, clusterLabels)
    print(f"Number of Clusters: {nClusters}, Silhouette Score: {silhouetteAvg:.4f}")
    y_lower = 10
    for i in range(nClusters):
        ithClusterSilhouetteValues = silhouetteValues[clusterLabels == i]
        ithClusterSilhouetteValues.sort()
        size_cluster_i = ithClusterSilhouetteValues.shape[0]
        y_upper = y_lower + size_cluster_i
        color = plt.cm.nipy_spectral(float(i) / nClusters)
        ax.fill_betweenx(np.arange(y_lower, y_upper), 0, ithClusterSilhouetteValues, facecolor=color, alpha=0.7)
        ax.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        y_lower = y_upper + 10

    ax.set_title(f"{nClusters} Clusters")
    ax.axvline(x=silhouetteAvg, color="red", linestyle="--")
    ax.set_xlabel("Silhouette Coefficient")
    ax.set_ylabel("Cluster Label")
    ax.set_yticks([])

plt.show()

Number of Clusters: 2, Silhouette Score: 0.4745
Number of Clusters: 3, Silhouette Score: 0.3905
Number of Clusters: 4, Silhouette Score: 0.3220
Number of Clusters: 5, Silhouette Score: 0.2930
Number of Clusters: 6, Silhouette Score: 0.2705

kmeans = KMeans(n_clusters=3, random_state=42, n_init='auto')
dataClusters = customerData.copy()
dataClusters['combined_cluster'] = kmeans.fit_predict(features)
customLabels = {
    0: "Good Debt Management, Good Payment Behaviour",
    1: "Good Debt Management, Bad Payment Behaviour",
    2: "Bad Debt Management, Bad Payment Behaviour"
}
dataClusters['combined_cluster'] = dataClusters['combined_cluster'].map(customLabels)

colours = cm.viridis(np.linspace(0, 1, 5))
fig = plt.figure(figsize=(10, 8))
ax = fig.add_subplot(projection='3d')

for cluster in range(3):
    clusterData = dataClusters[dataClusters['combined_cluster'] == customLabels[cluster]]
    ax.scatter(
        clusterData['debtManagementComponent'], 
        clusterData['paymentBehaviourComponent'], 
        clusterData['demographicFactorsComponent'], 
        color=colours[cluster], 
        label=customLabels[cluster], 
        s=20
    )

ax.set_title('Clustering on Combined Features', fontsize=14)
ax.set_xlabel('Debt management component', fontsize=12)
ax.set_ylabel('Payment behaviour component', fontsize=12)
ax.set_zlabel('Demographic factors component', fontsize=12, labelpad=-30)
ax.legend(title='Clusters', loc='upper left', fontsize=10)
plt.savefig('graphs/combinedClusters.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

clusterScoreDistribution = pd.crosstab(dataClusters['combined_cluster'], dataClusters['credit_score'], normalize='columns') * 100

clusterScoreDistribution.plot(kind='barh', color=['#d6e7e4', '#87bbae', '#48756a'], figsize=(10, 3))
plt.title('Credit Score Distributions In Combined Clusters')
plt.xlabel('Percentage')
plt.ylabel('Combined Cluster')
plt.legend(title='Credit Score', labels=['Poor', 'Standard', 'Good'])
plt.grid(axis='x', alpha=0.5) 
plt.show()

clusterScoreDistribution = clusterScoreDistribution.round(2)
annotations = clusterScoreDistribution.applymap(lambda x: f"{x}%")

plt.figure(figsize=(8, 3))
sns.heatmap(clusterScoreDistribution, annot=annotations, cmap='YlGnBu', cbar=False, fmt='', linewidths=.5, annot_kws={"size": 16})

plt.title('Credit Score Distributions In Combined Clusters', fontsize=16)
plt.ylabel('Combined Cluster', fontsize=14)
plt.xlabel('Credit Score', fontsize=14)
plt.xticks(fontsize=16)
plt.yticks(fontsize=16)
plt.savefig('graphs/combinedResults.png', dpi=300, bbox_inches='tight', transparent=True)
plt.show()

x = customerData[['debtManagementComponent', 'paymentBehaviourComponent', 'demographicFactorsComponent']]
y = customerData['credit_score']

xTrain, xTest, yTrain, yTest = train_test_split(x, y, test_size=0.2, random_state=42)

model = LogisticRegression(multi_class='multinomial')
result = model.fit(xTrain, yTrain)

yPred = model.predict(xTest)
accuracy = accuracy_score(yTest, yPred)
print("Accuracy: {:.2f}%".format(accuracy * 100))

Accuracy: 66.65%

classNames = [f"Credit Score {className}" for className in model.classes_]

coefficientsDF = pd.DataFrame(
    model.coef_.T, 
    index=x.columns, 
    columns= classNames)

coefficientsDF

oddsRatios = np.exp(coefficientsDF)
oddsRatios.columns = pd.MultiIndex.from_product([["Odds Ratios"], coefficientsDF.columns])
oddsRatios

	id	customer_id	month	name	age	ssn	occupation	annual_income	monthly_inhand_salary	total_emi_per_month	...	num_credit_inquiries	credit_mix	outstanding_debt	credit_utilization_ratio	credit_history_age	payment_of_min_amount	amount_invested_monthly	payment_behaviour	monthly_balance	credit_score
0	0x1602	CUS_0xd40	January	Aaron Maashoh	23	821-00-0265	Scientist	19114.12	1824.8434	49.574947	...	4	Good	809.98	26.822620	265	No	80.415300	High_spent_Small_value_payments	312.49408	2
1	0x1603	CUS_0xd40	February	Aaron Maashoh	23	821-00-0265	Scientist	19114.12	1824.8434	49.574947	...	4	Good	809.98	31.944960	266	No	118.280220	Low_spent_Large_value_payments	284.62915	2
2	0x1604	CUS_0xd40	March	Aaron Maashoh	23	821-00-0265	Scientist	19114.12	1824.8434	49.574947	...	4	Good	809.98	28.609352	267	No	81.699524	Low_spent_Medium_value_payments	331.20987	2
3	0x1605	CUS_0xd40	April	Aaron Maashoh	23	821-00-0265	Scientist	19114.12	1824.8434	49.574947	...	4	Good	809.98	31.377861	268	No	199.458070	Low_spent_Small_value_payments	223.45131	2
4	0x1606	CUS_0xd40	May	Aaron Maashoh	23	821-00-0265	Scientist	19114.12	1824.8434	49.574947	...	4	Good	809.98	24.797346	269	No	41.420155	High_spent_Medium_value_payments	341.48923	2

	count	mean	std	min	25%	50%	75%	max
age	100000.0	33.309240	10.764996	14.00000	24.000000	33.000000	42.000000	56.0000
annual_income	100000.0	50505.123439	38299.422075	7005.93000	19342.972500	36999.705000	71683.470000	179987.2800
monthly_inhand_salary	100000.0	4198.771620	3187.494355	303.64542	1626.761700	3096.378400	5961.745000	15204.6340
total_emi_per_month	100000.0	107.043617	130.035096	0.00000	29.197023	66.403020	146.827450	3776.0000
num_bank_accounts	100000.0	5.368800	2.593279	0.00000	3.000000	5.000000	7.000000	11.0000
num_credit_card	100000.0	5.533640	2.067073	0.00000	4.000000	5.000000	7.000000	11.0000
interest_rate	100000.0	14.532080	8.741330	1.00000	7.000000	13.000000	20.000000	34.0000
num_of_loan	100000.0	3.532880	2.446356	0.00000	2.000000	3.000000	5.000000	9.0000
delay_from_due_date	100000.0	21.068780	14.860104	-5.00000	10.000000	18.000000	28.000000	67.0000
num_of_delayed_payment	100000.0	13.334050	6.271409	0.00000	9.000000	14.000000	18.000000	28.0000
changed_credit_limit	100000.0	10.389303	6.789784	-6.49000	5.320000	9.400000	14.860000	36.9700
num_credit_inquiries	100000.0	5.772820	3.861137	0.00000	3.000000	5.000000	8.000000	17.0000
outstanding_debt	100000.0	1426.220376	1155.129026	0.23000	566.072500	1166.155000	1945.962500	4998.0700
credit_utilization_ratio	100000.0	32.285173	5.116875	20.00000	28.052567	32.305783	36.496664	50.0000
credit_history_age	100000.0	221.206880	99.681112	1.00000	144.000000	219.000000	302.000000	404.0000
amount_invested_monthly	100000.0	637.633413	2043.334171	0.00000	74.593977	135.952230	266.178185	10000.0000
monthly_balance	100000.0	403.120320	214.014558	0.00776	270.189028	337.114475	471.570645	1602.0405
credit_score	100000.0	0.888300	0.675120	0.00000	0.000000	1.000000	1.000000	2.0000

	customer_id	month	age	occupation	annual_income	monthly_inhand_salary	total_emi_per_month	num_bank_accounts	num_credit_card	interest_rate	...	changed_credit_limit	num_credit_inquiries	credit_mix	outstanding_debt	credit_utilization_ratio	payment_of_min_amount	amount_invested_monthly	payment_behaviour	monthly_balance	credit_score
0	CUS_0xd40	January	23	Scientist	19114.12	1824.8434	49.574947	3	4	3	...	11.27	4	Good	809.98	26.822620	No	80.415300	High_spent_Small_value_payments	312.49408	2
1	CUS_0xd40	February	23	Scientist	19114.12	1824.8434	49.574947	3	4	3	...	11.27	4	Good	809.98	31.944960	No	118.280220	Low_spent_Large_value_payments	284.62915	2
2	CUS_0xd40	March	23	Scientist	19114.12	1824.8434	49.574947	3	4	3	...	11.27	4	Good	809.98	28.609352	No	81.699524	Low_spent_Medium_value_payments	331.20987	2
3	CUS_0xd40	April	23	Scientist	19114.12	1824.8434	49.574947	3	4	3	...	6.27	4	Good	809.98	31.377861	No	199.458070	Low_spent_Small_value_payments	223.45131	2
4	CUS_0xd40	May	23	Scientist	19114.12	1824.8434	49.574947	3	4	3	...	11.27	4	Good	809.98	24.797346	No	41.420155	High_spent_Medium_value_payments	341.48923	2

	F-statistic	p-value	etaSquared
creditMixNum	14553.874261	0.000000e+00	0.278041
interest_rate	11940.754634	0.000000e+00	0.240106
paymentMinAmountNum	11041.741272	0.000000e+00	0.226116
num_credit_inquiries	9547.115407	0.000000e+00	0.201681
delay_from_due_date	8403.175426	0.000000e+00	0.181912
num_credit_card	7312.739885	0.000000e+00	0.162133
outstanding_debt	6878.150156	0.000000e+00	0.153982
num_bank_accounts	6709.186314	0.000000e+00	0.150769
num_of_delayed_payment	6427.953896	0.000000e+00	0.145368
num_of_loan	5778.281731	0.000000e+00	0.132624
changed_credit_limit	1917.378242	0.000000e+00	0.048287
monthly_balance	1755.573513	0.000000e+00	0.044393
annual_income	1099.083815	0.000000e+00	0.028262
monthly_inhand_salary	1083.676196	0.000000e+00	0.027877
age	627.651202	4.502314e-271	0.016337
total_emi_per_month	518.058897	3.452158e-224	0.013523
amount_invested_monthly	251.754293	1.064334e-109	0.006618
credit_utilization_ratio	82.541678	1.554879e-36	0.002179

	chi2	p-value	cramer's v
payment_of_min_amount	17090.736583	0.000000e+00	0.475492
credit_mix	32538.092101	0.000000e+00	0.463922
payment_behaviour	1553.138207	0.000000e+00	0.101036
type_of_loan	5180.361282	0.000000e+00	0.089370
occupation	160.750248	1.384363e-20	0.029634
month	125.971050	4.232041e-20	0.027216

	count	unique	top	freq
id	100000	100000	0x1602	1
customer_id	100000	12500	CUS_0xd40	8
month	100000	8	January	12500
name	100000	10139	Langep	48
ssn	100000	12500	821-00-0265	8
occupation	100000	15	Lawyer	7096
type_of_loan	100000	6261	['No Loan']	11408
credit_mix	100000	3	Standard	45848
payment_of_min_amount	100000	2	Yes	59432
payment_behaviour	100000	6	Low_spent_Small_value_payments	27588

	credit_score	loan_type	loan_count
0	2	Auto Loan	3566
1	2	Credit-Builder Loan	4101
2	2	Personal Loan	3659
3	2	Home Equity Loan	3802
4	2	Not Specified	3993
5	2	No Loan	3013
6	2	Debt Consolidation Loan	3574
7	2	Payday Loan	3897
8	2	Mortgage Loan	3845
9	2	Student Loan	3888
10	1	Auto Loan	16927
11	1	Credit-Builder Loan	18131
12	1	Personal Loan	17644
13	1	Home Equity Loan	17504
14	1	Not Specified	17529
15	1	No Loan	6137
16	1	Student Loan	16822
17	1	Debt Consolidation Loan	17522
18	1	Payday Loan	17962
19	1	Mortgage Loan	17225
20	0	Not Specified	13374
21	0	Auto Loan	13043
22	0	Student Loan	13698
23	0	Personal Loan	13361
24	0	Payday Loan	14101
25	0	Home Equity Loan	13286
26	0	No Loan	1130
27	0	Mortgage Loan	13418
28	0	Debt Consolidation Loan	13424
29	0	Credit-Builder Loan	13600

	customer_id	outstanding_debt	num_of_loan	num_credit_card	delay_from_due_date	num_of_delayed_payment	paymentMinAmountNum	age	annual_income	credit_score
0	CUS_0x1009	202.68	4	5	7	18	1	26	52313	1
1	CUS_0x100b	1030.20	0	4	13	7	0	19	113781	0
2	CUS_0x1011	473.14	3	3	27	14	1	44	58918	1
3	CUS_0x1013	1233.51	3	3	13	9	0	44	98621	1
4	CUS_0x1015	340.22	0	4	8	9	1	27	46951	2

	F-statistic	etaSquared
num_credit_card	1294.553660	0.194580
outstanding_debt	1228.892796	0.186552
num_of_loan	1006.234045	0.158095

	F-statistic	p-value	etaSquared
delay_from_due_date	1540.769126	0.0	0.223324
num_of_delayed_payment	1175.577320	0.0	0.179915

	F-statistic	p-value	etaSquared
annual_income	220.062583	2.179611e-94	0.039448
age	98.551351	3.877172e-43	0.018059

	Credit Score 0	Credit Score 1	Credit Score 2
debtManagementComponent	0.483378	-0.084296	-0.399082
paymentBehaviourComponent	0.410210	0.327788	-0.737998
demographicFactorsComponent	-0.015355	0.047711	-0.032357

	Odds Ratios
	Credit Score 0	Credit Score 1	Credit Score 2
debtManagementComponent	1.621543	0.919159	0.670936
paymentBehaviourComponent	1.507135	1.387894	0.478070
demographicFactorsComponent	0.984763	1.048868	0.968161